Credit constraints and business performance: evidence from public lending in Colombia

(1)

37

OCTUBRE DE 2014

Documentos CEDE

C E D E

ISSN 1657-7191 Edición electrónica.

Marcela Eslava

Alessandro Maffioli

Marcela Meléndez

(2)

Serie Documentos Cede, 2013-37 ISSN 1657-7191 Edición electrónica.

Octubre de 2014

Bogotá, D. C., Colombia

Teléfonos: 3394949- 3394999, extensiones 2400, 2049, 3233 [email protected]

http://economia.uniandes.edu.co

Ediciones Uniandes

Carrera 1ª Este No. 19 – 27, edificio Aulas 6, A. A. 4976 Bogotá, D. C., Colombia

Teléfonos: 3394949- 3394999, extensión 2133, Fax: extensión 2158 [email protected]

Edición y prensa digital: Cadena S.A. • Bogotá Calle 17 A Nº 68 - 92 Tel: 57(4) 405 02 00 Ext. 307 Bogotá, D. C., Colombia www.cadena.com.co

Impreso en Colombia – Printed in Colombia

(3)

C E D E

Centro de Estudios sobre Desarrollo Económico

Credit constraints and business performance: evidence from public

lending in Colombia

Marcela Eslava, Alessandro Maffioli, Marcela Meléndez*

Abstract

Whether public lending to firms effectively eases credit constraints has been widely studied for very small businesses. The evidence documented for larger firms refers to lending that is significantly subsidized and targeted to these businesses, so the estimated positive effects may reflect poor allocation of public credit. This paper investigates the impact on its beneficiaries of a wide, untargeted and unsubsidized, lending program in Colombia. We use data on all non-micro manufacturing firms and all formal credit operations. After correcting for potential selection biases using econometric techniques, we find that Bancóldex loans increase firms’ employment, purchases of inputs, investment, and output for small (but non-micro) firms, while large firms experience increases in variable inputs, but not on investment. While both short-term and long-term Bancóldex loans are found to have positive impacts on output, input demand and employment, only long-term loans increase investment. Moreover, short-term loans have a larger impact on

input demand than long-term loans. Our findings also indicate that Bancóldex’

beneficiaries end up with improved overall credit conditions after receiving Bancóldex credit: the amount of credit received goes up, the duration of the loans increases, and

beneficiaries are able to establish credit relationships with more financial intermediaries.

Though the interest rates go down, in this dimension the effect is small.

Keywords: Credit constraints,public development banks, firm growth.

JEL Classification: G28,H43, L25, O12, O54

*_{Eslava: Department of Economics, Universidad de Los Andes ([email protected]). Maffioli:}

Inter-American Development Bank (IDB) ([email protected] elendez: Econ Estudio

([email protected]). We would like to thank Juan Sebastián Galán and Laura García for excellent research assistance. We are extremely grateful to Bancoldex, DANE, and Superintendencia Financiera for their constant advice, their willingness to provide access to data, and for their efforts to establish the mechanisms that made such access possible within the strict reserve requirements that protect the respective data. We also thank the financial support of the IADB for an initial stage of this

Credit constraints and business performance: evidence from public

lending in Colombia

Abstract

([email protected]). We would like to thank Juan Sebastián Galán and Laura García for excellent research assistance. We are extremely grateful to Bancoldex, DANE, and Superintendencia Financiera for their constant advice, their willingness to provide access to data, and for their efforts to establish the mechanisms that made such access possible within the strict reserve requirements that protect the respective data. We also thank the financial support of the IADB for an initial stage of this study. Any remaining errors are our own. The views here expressed do not represent those of the Inter-American Development Bank.

Credit constraints and business performance: evidence from public

lending in Colombia

Abstract

(4)

2

Restricciones crediticias y desempeño empresarial: evidencia de un

programa de crédito público en Colombia.

Marcela Eslava, Alessandro Maffioli, Marcela Meléndez

Abstract

Investigamos restricciones crediticias al desarrollo empresarial, aprovechando un programa de crédito público en Colombia. Usando microdatos de gran riqueza sobre establecimientos manufactureros y operaciones crediticias formales encontramos que: 1) Los beneficiarios del programa de crédito público, especialmente las empresas pequeñas, incrementan su demanda por insumos, su compra de activos fijos y su producto de manera significativa; 2) El efecto se concentra en beneficiarios de líneas de largo plazo; 3) Los beneficiarios también terminaron por incrementar su nivel total de crédito y el número de intermediarios financieros de los que obtuvieron crédito. Un modelo sencillo indica que estos resultados son consistentes con la hipótesis de que las empresas en Colombia enfrentan restricciones significativas en su acceso al crédito, en particular en lo que se refiere a crédito de largo plazo. Las empresas pequeñas se ven afectadas de manera particularmente fuerte.

Palabras clave: restricciones de crédito, banca de desarrollo empresarial,

crecimiento empresarial.

(5)

1. Introduction

Problems in access to credit are widely seen as a main obstacle for productive growth, while lending by public development banks has been proposed as a solution: public banks could channel funds to productive activities that, even if promising, may not flourish because of lack of credit access. But despite the salience of both credit market imperfections and public development banks in the public discussion, there is little microeconomic evidence about how both actually impact firms. Much of the evidence we have about the prevalence and effects of credit constraints and public development

banking comes from cross-country or cross-industry studies.1_{Micro econometric analyses}

of public lending have been scarce, and many concentrate on the question of whether the

allocation of public lending is politicized.2

This study aims at improving our understanding of credit constraints by analyzing the effect of lending by Bancóldex, Colombia’s main development bank, on beneficiary firms. Two different dimensions are analyzed: firm performance, and credit conditions. In particular, we use data on all non-micro manufacturing establishments in the country to study how the firm’s output and demand for inputs change after receiving Bancóldex credit, relative to non-recipients. We then use data on all commercial- and microcredit operations of the supervised financial system to analyze how conditions such as interest rates on loans, loan maturities, and number of banks that lend to a firm, change after a firm has received Bancóldex credit, relative to non-beneficiaries. Our period of analysis is 2004-2009. A simple model of lending by firms is used to interpret the implications of our findings about the extent of credit constraints; the role of quantity vs. price effects of public lending; and more generally the mechanisms by which public lending can relax credit constraints.

1_{This evidence indicates that growth is slower in countries with low financial depth and in sectors with}

high depe ndence on external financing, but also in countries where public banks play a more prominent role. See, e.g., Galindo and Micco (2004); La Porta, Lopez de Silanes, and Shleifer (2002); Barth, Caprio and Levine (1999); Beck and Levine (2002).

2_{See Carvalho (2014); Dinç ( 2005); Cole (2009); Micco et al. (2007); and Sapienza, 2004. Their findings}

(6)

Other recent micro econometric studies have shed light on the presence and extent of credit constraints. McKenzie and Woodruff (2008) and De Mel, McKenzie and Woodruff (2008) examine the effect of randomly allocating grants to microenterprises in Mexico and Sri Lanka, respectively. They find evidence consistent with these firms facing tight credit constraints. Banerjee and Duflo (2014) examine the effect of subsidized credit by one particular public bank in India on the growth of beneficiaries. They compare the growth of beneficiaries of subsidized public credit to the growth of beneficiaries of unsubsidized public credit. They estimate the effect by taking advantage of a natural experiment that changed conditions for access to the subsidized loans. They also find evidence consistent with credit constraints being an important obstacle for the growth of these businesses.

Our study complements those previous pieces in several manners. First, we add the dimension of credit conditions to that of firm performance. To the extent of our knowledge, this is the first direct econometric assessment of how public lending affects the conditions of credit that beneficiary firms face. Second—as in Banerjee and Duflo (2014)—our findings are not limited to microenterprises, which typically represent a small chunk of business activity and for which there is little room for controversy about the importance of credit constraints. Our study, in fact, covers a wide population of firms and a very large fraction of non-agricultural recipients of public credit to firms. In that sense, our findings are more widely applicable to other contexts. The cost, as is frequently the case, is a less stark identification of effects than the one obtained in experimental settings.

(7)

about politicized allocation of public lending (arguably eliminating them) and pushes us to

understand the effect of public lending in a setting where it is not subsidized.3

After correcting for selection biases using econometric techniques, we find that using Bancóldex loans increases firms’ employment, purchases of inputs, investment, and output. Most of these effects are concentrated on small firms that receive Bancóldex loans. While both short-term and long-term Bancóldex loans are found to have positive impacts on output, input demand and employment, only long-term loans increase investment. Moreover, short-term loans have a much larger impact on input demand than long-term loans. Our findings also indicate that Bancóldex beneficiaries end up with improved overall credit conditions after receiving Bancóldex credit: the amount of credit received goes up, the interest rates go down, the duration of the loans increases, and beneficiaries are able to establish credit relationships with more financial intermediaries.

Our results contribute to the understanding of credit constraints in middle-income countries. They are consistent with Colombian firms facing obstacles in their access to credit, particularly important for projects that require long-term lending. A simple conceptual framework (Section 3) shows that an increase in firms’ output and input use after using public lending (but not fully substituting private lending) can be interpreted as a signal that the firm was previously credit constrained, and consequently is not simply expanding profits by using public lending to replace available, though more expensive, alternative funding sources. This is also consistent with the finding that, while accompanied by only a mild decrease in the annual interest rate of 2 percentage points (with an average interest rate of 23 pp and a standard deviation of 9 pp), Bancóldex use shoots credit to the firm by about 50%.

The paper is organized as follows: Section 2 describes the institutional environment and the data; Section 3 presents a conceptual framework and reviews previous studies. Section 4 discusses our empirical approach. Section 5 presents the results of our study, and Section 6 concludes.

(8)

2. Institutional Background and Data a. Institutional Background

Bancóldex started operating in its current form after 2003.4_{Over our period of study}

(2004-2009), its operations focused on second-tier lending, though they also included

training services, with the latter restricted to micro enterprises.5_{The present study focuses}

exclusively on Bancóldex’ credit operations The second-tier model implies that funds are lent to intermediary financial institutions, which then lend those funds, at higher rates, to final beneficiaries. Since the intermediary institution takes on the risk of default, Bancóldex funds are subject to only moderate risk. In fact, as of December 2013 Bancóldex’ bad credit was at 0.001% of its total credit (Bancóldex, 2014). Intermediaries are usually commercial banks, who have the incentive to carefully screen applicants and are, in principle, not subject to the type of political considerations that many have pointed as a source of inefficiency for direct public loans (e.g. Dinç, 2005; Cole, 2009; Micco et al., 2007; Sapienza, 2004). Almost all supervised financial institutions in fact intermediated Bancóldex credit over our period of study.

In general, Bancóldex credit is not subsidized, not targeted to specific types of producers, and not subject to other benefits. Bancóldex has a mandate to provide credit for banks to serve firms in all sectors and of all sizes, and to engage only in profitable credit operations. The only targeted credit lines are those called special lines, which amount to less than 5% of credit value over our period of analysis. Special lines are usually funded by specific local governments, restricted to the respective region, and frequently targeted to specific types of firms within the region. They are also generally restricted to a specific period of time.

Bancóldex’ potential impact on firms is particularly intriguing given the absence of strategies to target most of Bancóldex funding to specific firms, or to provide special benefits to recipients. Is it necessary for a development bank to provide subsidized or targeted credit

4_{Its current-day operations were previously divided between credit for exporting, in charge of Bancóldex}

itself, and more general development credit, in charge of IFI (which disappeared in 2003).

5_{Starting in 2010—that is, after our period of study—Bancóldex was put in charge of}

(9)

in order to alleviate credit constraints faced by firms? This is partly what the current study

aims at responding.6

2.2. Data

One fundamental feature of this study is the use of detailed data on each credit recipient, as well as for counterpart non-recipients, to evaluate the effects of government-funded credit. With very few exceptions, on the side of firm performance previous studies have lacked access to this level of detail and have therefore used highly aggregated data to infer the effect of government-owned banks. Moreover, we go beyond the effect of public loans on firm performance to also characterize the effect on the conditions of credit that the business faces. This angle is novel.

Our performance analysis focuses on how obtaining credit from Bancóldex affects a firm’s output, use of labor, purchases of materials, and investment. To this end, we use information from the Annual Manufacturing Survey (AMS). The AMS database, developed

and owned by the National Statistical Agency DANE (Departamento Administrativo Nacional

de Estadistica), provides annual information on all manufacturing establishments with 10 or

more employees, and allows tracking each establishment over time. 7_{The unit of}

observation in our performance analysis is thus the establishment rather than the firm,

though one should note 97% of firms in the Survey own a single establishment.8

The AMS provides information on production; use of labor; purchases of inputs and purchases of fixed assets; and interest payments. We use purchases of fixed productive assets, in particular machinery and buildings, as our definition of investment. All values are expressed in pesos of 2009 using the CPI as a common deflator. Location and sector of activity, at a three-digit level are also provided. Bancóldex beneficiaries are flagged in the

6_{In terms of its funding, Bancóldex obtains funds from Term Deposit Certificates (50% of liabilities);}

bonds (20%), credit from multilaterals (about 25%), and credit from other sources. (Participations as of December of 2013, according to Bancóldex, 2014). Therefore, Bancóldex does not have a clear advantage compared to private banks in terms of the cost of its funding. In fact, its short-term funding is more costly than that of standard banks, given that it cannot rely on checking or savings accounts. Its long-term funding seems to be similar in cost to other sources used by private banks (Vargas, 2014)

7_{Establishments with fewer employees are included in the survey if they either belong to firms that have}

assets above 500 minimum monthly wages or have other establishments with 10 or more employees.

8_{In any case, all of our estimations control for a dummy equal to one for establishments belonging to}

(10)

version of the data that we use, together with the amount of Bancóldex credit, and the

agreed maturity of the loan.9_{All establishments belonging to a firm that received}

Bancóldex credit are marked as beneficiaries.

The kind of rich information the AMS provides for the manufacturing sector is not available for other sectors of the economy. Our analysis of performance thus has the limitation of covering manufacturing only. Still, a large chunk (about 25 percent) of the

Bancóldex beneficiaries with 10 or more employees is covered by the survey.10_While

businesses with fewer than 10 employees represent the most numerous group of Bancóldex beneficiaries, they receive less than 20 percent of Bancóldex’ credit disbursements during the observed period (see below). Even with the limitation of focusing on just one major activity, the AMS is also a particularly valuable source given its census-type coverage of establishments over nine employees: firm-level surveys in other countries frequently cover only random samples of small and medium-size businesses.

With regards to access to credit from the financial sector, we use information on all credit operations intermediated by formal financial intermediaries from the beginning of 2004 through 2009. These data are housed at the Colombian Financial Superintendence, the agency that oversees the activities of all formal financial intermediaries, including all banks. All institutions supervised by the Superintendence are required to provide information on all financial transactions. The database we use contains annual information, as of the last quarter of each year, on each outstanding credit operation, detailing outstanding balance, date of disbursement, interest rate and maturity initially agreed upon, and use of collateral. We have information on all microcredit and commercial credit

operations, to firms in all sectors of the economy.11_{. The version of the data we use has}

been aggregated at the firm level, for which we observe total credit; number of loans; number of individual financial institutions that provided these loans; average interest rate;

9_{We thank Bancóldex and DANE for agreeing on terms that allowed flagging Bancóldex beneficiaries in}

EAM data.

10_{The distribution of Bancoldex loans in 2000-2007 received by firms of 10 or more employees is as}

follows: wholesale and retail trade, 32 percent; manufacturing, 25 percent; nonfinancial services, 23 percent; transport, storage and communications, 14 percent; and other sectors, 6 percent.

(11)

and average loan duration. Bancóldex beneficiaries are marked in the version of the data we use, in each year in which they received Bancóldex credit.

Ideally, we would have merged the two datasets, the AMS and the Superfinanciera, but confidentiality restrictions prevent us from doing this. Therefore, the two analyses proceed separately. We would have also ideally dropped microenterprises from the dataset used to conduct our analysis relating credit conditions, to make it more consistent with the performance analysis, but the Financial Superintendence data has no record of the size of credit recipients.

2.3. Bancóldex in numbers

Table 1 shows the evolution of Bancóldex financing activity by the size of the beneficiary. The source is Bancóldex’ own records, which contain a categorical variable

classifying loan recipients by their size. 12_{The lower panel of the Table reproduces the}

upper one restricting the sample to loans intermediated by formal financial institutions, again as reported by Bancóldex.

Though most loans are directed to micro beneficiaries, as mentioned above they represent less than 20% of total value over the period. Most Bancóldex credit is intermediated by the formal financial sector, particularly for non-micro firms (the only ones included in our performance estimations). Bancóldex lending activity has been growing over time. In 2009 there was a discrete jump in the number of loans to micro enterprises intermediated by the supervised financial sector. Because of the very small size of these loans, the extreme increase in numbers was not reflected in a proportionally large increase in the fraction of value represented by lending to micro firms. To avoid biases related to this change in the composition of loans, our baseline estimations regarding credit conditions (based on data for the supervised financial sector) exclude 2009, though we also check the robustness of findings to keeping 2009 in the sample. A similar adjustment is not necessary in our performance estimations, as micro establishments are not covered in that part of the analysis.

12_{Size categories in Table 1 are based on assets held by the firm, as reported in its loan application. The size}

(12)

To assess the relative importance of Bancóldex in the provision of credit in the country, we compare Bancóldex to non-Bancóldex credit using the data from the Financial Superintendence. Table 2 summarizes several features of bank-firm relationships with at

least partial Bancóldex funding, compared to features of non-Bancóldex relationships.13

Over the period of our analysis, lending funded from Bancóldex represented about 5% of all credit operations in the country. That is, it is far from being negligible as a single source of funding, but it is also far from dominating total credit in the country. These numbers also reveal the predominance of privately-funded credit in the country: though there are other public banks, Bancóldex is by far the largest.

Table 2 also reveals other interesting patterns. First, in terms of numbers of relationships, Bancóldex’ participation is greater in the segment of commercial credit than

that of microcredit up to 2009. Consistent with Table 1, however, this is reversed in 2009.14

Bancóldex credit is on average cheaper than average private credit, with similar maturities. But, these patterns are dominated by commercial credit. Microcredit by Bancóldex, in fact, tends to be shorter term and more expensive than microcredit funded using other sources. These differences may be related to banks using Bancóldex credit to serve different segments of the market, a potential selection that we try to address in the empirical section using econometric techniques.

3. Conceptual framework

We motivate our empirical analysis using a simple static model of credit for working

capital, in the spirit of that presented by Banerjee and Duflo (2014).15_{Consider a firm that}

produces revenue using a vector of variable inputs , generating profits

13_{Because of the structure of the data, we compare relationships rather than loans. A given loan is either}

totally funded by Bancóldex or totally funded from other sources. But, a firm can have more than one loan with a given bank, some of those funded from Bancóldex and other from other sources. Still, over 90% of relationships in our database, feature a single loan.

14_{Care must be exercised when comparing Table 1 to Table 2, however, as Table 1 classifies loans according}

to the size of the recipient while Table 2 classifies them according to the size of the loan.

15_{As will become clear, the model does not literally represent a static situation (in which credit would not}

(13)

where is a (row) vector of factor prices, and is assumed increasing and strictly

concave in . Inputs must be paid out of working capital; that is, inputs must be paid for

before revenue is realized Working capital may come from the firm’s internal resources— profits from previous periods, social capital, etcetera—or from credit, where a firm with perfect access to credit is defined as facing the same cost of internal and external financing,

and therefore can fund any level of that generates benefits greater than the opportunity

cost.

The firm takes into account the cost of these funds, either external or internal or both, when making its optimal decision about input use. Denoting the opportunity cost of using internal funds as r, the problem of the firm in absence of credit imperfections can be written as:

The optimal demand for inputs by this perfectly unconstrained firm is given by

associated with an optimal production level of This unconstrained firm may be

paying for its optimal input bill, using internal funds, external funds, or a combination

of both, since it is indifferent between the two sources.

Suppose now that the firm faces credit rationing at rate —the amount of funds it

can access at cost r falls short of what it would demand at that rate—and can only access

additional credit at a higher rate . One possible scenario is represented in Figure 1,

denoting as and the maximum levels of funds available to the firm, respectively, at

rates and . ̂ in the Figure is the maximum amount of funds the firm is willing to

borrow at rate ; clearly, ̂ . With this particular distribution of and in relation

to and ̂, the firm purchases inputs up to , with the first part of the bill

(14)

Higher (potentially infinite) costs of external funds compared with internal ones capture the imperfections in credit markets, which imply a reduction in firm size from --the level of activity that would be efficient—to ̂. The size of this negative effect on production grows with the extent of the wedge between the cost of internal and external funds.

Clearly, the potential role for credit from a public bank in addressing these effects of credit market imperfections on economic activity depends on the amount of funds it is willing to lend to the firm, and the rate at which it offers those funds, compared to and

. Assuming that lies above + , the different cases are summarized in Result 1 (and later illustrated in Figure 2), where we denote the rate at which public credit is offered as , and the maximum amount of credit that the firm is willing to borrow at rate as

̂:

Result 1: An amount of public credit made available to the firm at a rate

expands the firm’s production if:

i. and ̂. That is, if public credit is cheaper than market credit and, in the absence of public credit, the firm is credit rationed at rate .

f’(X)

𝑟𝑟

𝑋𝑋

𝑖𝑖

𝑋𝑋

𝑖𝑖

𝑋𝑋

𝑚𝑚

𝑟𝑟

𝑚𝑚

𝑋𝑋

(15)

ii. , ̂. That is, if public credit is more expensive than public credit and, in absence of public credit the firm is credit rationed at

rate (with ).

iii. and, even though ̂ the amount is sufficiently large that it fully substitutes market lending. That is, the firm is

not rationed at rate , but the intervention fully substitutes market

lending at a cheaper rate.

If, on the contrary, the firm’s access to credit at market or higher rates is not rationed, and public lending offered to the firm is not large enough and not cheap enough to fully substitute market credit, then public credit would partially substitute market credit, increasing firms’ net profits, but would not increase firm’s

input demand and production.16

The intuition behind this result is shown in Figures 2.i. through 2.iii. which represent items i. through iii in Result 1, respectively. In Figure 2.i., market credit is initially rationed

( ̂) so that public credit offered at or below market rates is accepted. The firm

uses all of its internal resources, complements them with public credit, and may continue to

use some market credit if ̂. Total credit goes from to { {

̂} { ̂}} . In Figure 2.ii., the initial rationing at market rates is sufficiently strong that the firm uses public credit to complement market credit, even though the cost of public funds exceeds that of market funds. Total credit goes from

to { ̂}. Finally, in Figure 2.iii. there is no credit rationing at market rates, but the amount of relatively cheap public credit is sufficiently large that the firm is induced to demand more credit, fully substituting market for public credit. Total

credit goes from to { ̂}.

16_{We ignore above the possibility that public resources are lent at rates below r because the rules governing}

Bancóldex are such that it cannot lend at a loss, and it is not clear how the bank could obtain funding at rates below firms’ opportunity cost If we allowed this possibility, then public credit would clearly expand the firm’s demand for inputs Assuming that the firm’s opportunity cost reflects the socially best

(16)

f’(X)

𝑟𝑟

𝑋𝑋

𝑚𝑚

𝑋𝑋

𝑖𝑖

𝑟𝑟

𝑚𝑚

𝑋𝑋

Figure 2i: Demand and supply for funds with public credit, case i.

𝑟𝑟

𝑝𝑝

f’(X)

𝑟𝑟

𝑋𝑋

̂

𝑝𝑝

𝑟𝑟

𝑚𝑚

𝑋𝑋

Figure 2ii: Demand and supply for funds with public credit, case ii.

(17)

These results have important implications that we use to guide the interpretation of our empirical analysis. First, unless public credit is offered in sufficiently large amounts and at a sufficiently low interest rate to fully substitute the firm’s use of market credit, a positive effect of public credit on firm performance is indicative of the firm being credit constrained, in the precise sense of being credit rationed—in absence of public credit—at the highest interest rate possible (Banerjee and Duflo, 2014). (Moreover, even in the exceptional case where public credit to the firm is sufficiently large and cheap to fully substitute market credit, public lending leading to a firm expansion indicates credit market imperfections in the form of rationed credit at the firm’s opportunity cost Second, public credit may mitigate the negative effects of credit rationing on firm performance for various reasons: (1) A pure quantity effect: maintaining the highest rate at which the firm lends, it increases the amount of resources the firm lends (item i in Result 1, with ̂). (2) A quantity effect with increase in rates: the firm is sufficiently constrained to start with, that public credit increases its credit use even if offered at a cost above market rates (item ii in Result 1). (3). A price-reducing effect: a sufficiently large amount is offered at below market rates, that the firm ceases to use market credit. The highest rate at which the firm lends goes down, inducing an increase in its demand for credit (item iii in Result 1, or item i with

f’(X)

𝑟𝑟

𝑋𝑋

𝑚𝑚

𝑋𝑋

𝑖𝑖

𝑟𝑟

𝑚𝑚

𝑋𝑋

𝑖𝑖

𝑋𝑋

𝑚𝑚

Figure 2iii: Demand and supply for funds with public credit, case iii.

𝑟𝑟

𝑝𝑝

𝑋𝑋

𝑝𝑝

(18)

̂). We explore these possibilities in the empirical analysis, by examining the impact of Bancóldex loans not only on firms’ input demand and output, but also on credit and interest rates.

One additional observation is useful before proceeding. Though this conceptual framework is best suited to understand the use of credit for working capital, it can still plausibly accommodate public credit for investment in fixed capital. The main difference between investment and purchases of variable flexible inputs is the lumpy character of investment. But suppose we collapse all of the complexities that arise in the context of dynamic decisions subject to non-convexities to the assumption that investment can only be undertaken in chunks, and that the purchase of a desired machine costs an amount I* in Figure 3. Then public credit expands investment, and subsequently output, only if provided in an amount sufficiently large that the firm can borrow from the public bank

at a cost not above . Moreover, in particular cases (e.g. Figure 3 with ) the availability of public credit also induces the firm to use private credit that would not be accepted without complementary credit funding, as using the two sources is necessary to make possible a level of investment (I*).

f’(X)

𝑟𝑟

𝑋𝑋

𝑖𝑖

𝑋𝑋

𝑖𝑖

𝑋𝑋

𝑚𝑚

𝑟𝑟

𝑚𝑚

𝑋𝑋

𝐼𝐼

Figure 3: Demand and supply for funds, indivisible inputs

(19)

Motivated by these implications of the conceptual analysis, we take advantage of the rich longitudinal data on both recipients and non-recipients of Bancóldex credit to address a series of questions: 1) does receiving a loan from Bancóldex improve firm performance, thus indicating the presence of credit market imperfections, likely even credit constraints, for firms in our sample?; 2) do we see evidence that these effects are stronger for smaller than for larger firms, where the latter are less likely to be rationed in their access to internal funding (i.e. funding at rate r)?; 3) is Bancóldex credit generally associated with lower interest rates?; 4) to what extent Bancóldex credit substitutes/complements/induces the use of other types of credit by the firm? Other questions not borne directly by the above conceptual analysis are also relevant. In particular, we investigate potential heterogeneous effects of Bancóldex credit according to loan maturity, besides those relative to firm size.

4. Empirical Approach

We attempt to identify whether firms that received credit from Bancóldex increased their performance more or saw more marked changes in the conditions of their credit, compared to firms that arguably had similar access to credit but did not benefit from Bancóldex lines. Our baseline independent variable is a dummy indicating whether the firm received credit from Bancóldex in the respective year. We estimate equations of the form:

(1)

where is a dummy indicating whether the firm had loans funded by Bancóldex in year t,

is a vector of control variables, is a vector of year dummies, and is a random

error term. When estimating the effect of Bancóldex on performance, is one of three

alternative measures of factor demand and output: labor, input consumption, investment, and output. When, instead, we estimate how Bancóldex affects the conditions of credit

faced by the firm, is, alternatively, the firm’s credit balance; the number of financial

intermediaries from which it received loans in the period; the average interest rates on its loans; or the average maturity of those loans.

A set of control variables, , is introduced in our specifications. These capture

(20)

estimations, conducted using Manufacturing Survey data, we include year dummies; a dummy for establishments belonging to multi-establishment firms; age and age squared; and a dummy for whether the firm paid interest over financial liabilities in the previous year. The latter is intended to control for previous access to overall credit, very crudely given the impossibility to bring together the Manufacturing Survey data and the data on loans from the Financial Superintendence. In turn, in estimations for credit conditions, which use the Superintendence data, we include year dummies; controls for whether the firm had only commercial loans, or had at least one loan classified as micro-credit in the respective year; the percentage of loans to the firm on which a guarantee was required in that year; and the maximum number of loans that the firm received from a single bank in the period (in logs), to proxy for firm size.

As is always the case when estimating the effect of a given program on its beneficiaries, a major source of concern is the possibility that firms that receive Bancóldex credit are selected according to characteristics that are also correlated with the outcome variables we consider. It is difficult to a priori establish what those characteristics may be: there is no explicit criterion according to which Bancóldex gives credit to firms. In fact, the allocation of its loans is not even in the hands of Bancóldex, as it is the intermediary banks to assign Bancóldex funds to loans. There is similarly no clear unique criterion for intermediary banks to use Bancóldex funds rather than other funds to finance a loan to a given firm, nor does it seem to be the case that firms decide themselves whether to take a loan funded by Bancóldex or by the bank’s own resources—so that it is not clear that self-selection should be an issue--. Interviews conducted by Vargas (2014) with heads of commercial departments at most banks in the financial system indicate that a bank’s decision to use Bancóldex funding for particular loans has more to do with conditions faced by the bank than with characteristics of the loan recipient.

Still, it may well be that firms with given characteristics are more likely assigned

Bancóldex’ funded loans, biasing OLS estimates of in equation (1), though it is difficult

(21)

Mel, Mc Kenzie and Woodruff, 2008), we use econometric techniques to address concerns about potential endogeneity.

In the end, we want to compare Bancóldex beneficiaries to firms that faced similar conditions in accessing credit but did not benefit from Bancóldex funding. Because access to credit is not directly observable, we create these comparison groups by using fixed effect estimators, either alone or in combination with propensity score matching techniques, where the latter help us equate treatment and control groups in terms of observable characteristics in the year prior to receiving Bancóldex credit.

In particular, we use two different estimation approaches. The first approach includes plant- or firm-level fixed effects (depending on the level of observation in the

respective set of data).17_{This approach has the advantage of controlling for unobservable}

characteristics, but the problem that it assumes all those characteristics to be fixed over time for a particular firm. We then complement this estimation strategy with another that restricts the dataset to beneficiaries and their “nearest neighbors”, identified using propensity score techniques. The propensity score—predicted probability of receiving Bancóldex credit in a given year—is estimated on the basis of a set of characteristics in the previous year. Each beneficiary is then matched to a firm that did not receive Bancóldex credit (over our whole period of observation), whose propensity score in the respective

year was closest to that of the beneficiary among the group of non-beneficiaries.18_This

generates a restricted dataset for each year, including that year’s beneficiaries and their nearest neighbors. A panel is finally constructed including the full history of each beneficiary and nearest neighbor. It is over this end panel that our estimations are conducted for our second estimation approach.

Propensity scores are established by estimating models where the probability of receiving Bancóldex in a given year is a function of different performance variables lagged

17_{The unit of observation in the Manufacturing Survey is the plant, while the unit of observation in the}

Financial Superintendency data is a firm.

18_{We match to a single nearest neighbor although less strict matching algorithms are obviously available,}

(22)

one period;19_{the first lags of the control variables of the baseline regression (excluding,} obviously, the time effects); and, in the performance estimations, fixed effects by four-digit sector, location, and type of legal organization. A separate participation model is estimated in each year, and a nearest neighbor is found for each Bancóldex beneficiary in that particular year. Equation (1) is then estimated over a set of data that pools across years and includes all firms that received Bancóldex in at least one year, and all firms that were identified as nearest neighbors to those beneficiaries, with their respective full histories.

Our baseline estimation of equation (1) is also extended to examine potential heterogeneous effects. In particular, we differentiate the effect of long- and short-run Bancóldex loans; and the effect of Bancóldex on larger vs. smaller establishments. We also examine alternative timings for the effect of Bancóldex credit. Robustness tests have been conducted including lagged dependent variables rather than fixed effects in the different specifications, and restricting the matching to firms that received Bancóldex in 2004 for the first time and their nearest neighbors in that year, rather than the full 2004-2009 database.

5. Results

5.1. Bancóldex and firm performance

We begin by discussing our findings regarding the effect of Bancóldex credit on firms’ demand for inputs and output. This estimation uses data from the Annual Manufacturing survey, covering all non-micro manufacturing establishments in the country. Descriptive statistics for this estimation are presented in Table 3. 11% of all observations correspond to Bancóldex beneficiaries.

Results of estimating equation (1) are presented in Table 4. For each outcome under consideration, two adjacent columns present results using the two estimation strategies described in Section 4: a FE estimation, and a FE estimation conducted over the dataset restricted to include only beneficiaries and their nearest neighbors in terms of the probability of receiving Bancóldex (marked as FE+PSM in our tables). The different vertical panels present different specifications for the treatment variable. Figure 4

19_{The set of lagged characteristics included in the models covers output, TFP, exports, and investment for}

(23)

summarizes the results of the propensity score matching approximation, depicting the distribution of propensity scores in both the unrestricted and the restricted datasets. Restricting to beneficiaries and their nearest neighbors eliminates pre-existing differences in these scores. At an individual level, pre-existing differences in each of the outcomes of interest also become negligible after the restriction (Appendix A1).

The use of Bancóldex is associated with an increase in the demand for inputs and a related increase in output (Panel A). Within plants, employment increases around 4% for Bancóldex beneficiaries in relation to non-beneficiaries, input demand increases above 5%, output increases a little below 5%, and investment grows markedly by about 20%. Investment comes in spikes (Haltiwanger and Cooper, 2006; Doms and Dunne, 1998), so that, contingent on part of the credit being used for purchases of fixed capital, large investment effects are not surprising. Interestingly, results are very similar after restricting

the database using propensity score matching.20

The positive effects of Bancóldex credit on employment and output exhibit persistence over time (Panel B). In fact, though the coefficient associated with the treatment variable decays over time, it does so at a very slow pace. By the third year, the effect over employment is still above 2.5%, and that over output remains in a similar neighborhood. Similar persistency is not found for the effect on investment. Although the coefficient for the current Bancóldex dummy is still positive and large (but imprecisely estimated), the one on the lagged treatment becomes negative. This is probably related to the well-known lumpy behavior of investment, which implies that a large investment episode today reduces the probability that the firm invests tomorrow. The effect on material purchases exhibits only moderate persistence: the coefficient halves by year two and becomes insignificant by year three. Together, these results suggest that Bancóldex loans are at least partly used to undertake investments that have persistent effects on the operation of plants. In light of the conceptual framework discussed in section 3, they also

20_{This is consistent with the view that intermediaries, once having dedided to grant credit, choose whether}

(24)

suggest that these firms are rationed in their access to private credit, in a way that prevents

them from undertaking investments and growing to their optimal size.21

Table 5 shows the results of re-estimating equation (1) separately for small and large establishments (excluding medium establishments). Our conceptual framework suggests that large establishments should see a lesser impact from public credit on their use of inputs and output if, as is plausible, they have more internal financing and are less rationed with respect to external financing. Results in Table 5 are broadly consistent with this view. Following Colombian legal definitions, small establishments are those with 50 or less employees, and large establishments those with 200 or more employees. For employment, output, and investment, the coefficients estimated for large establishments are about half the size of those for small establishments. They are also imprecisely estimated, though in some of our robustness proofs do turn out to be statistically significant. Interestingly, however, the effect of Bancóldex on material purchases is particularly large for large

establishments.22_{That is, small establishments seem to use Bancóldex funding relatively}

more intensely for funding investments, while large establishments seem to use it mainly for working capital, and especially at times of particularly large working capital needs.

Table 6 presents results of estimating separately the effect of long- and short-run Bancóldex loans. Firms are more likely to face credit rationing from banks for long run projects. The reasons are varied: long run projects also tend to be large projects and may thus require co-financing by several banks (Dewatripont and Masking, 1995; Armendáriz, 1999); there may also be greater uncertainty about their returns; moreover, the deposits from which private banks frequently fund loans are mostly short run.

In our estimations, long-run loans are defined as those with maturity of at least three years, short-run ones as those with maturity of up to eighteen months. Establishments that obtained other types of Bancóldex credit are left out, so that the control group corresponds

21_{Note, from section 3, that this interpretation holds as long as Bancóldex credit is not offered at subsidized}

rates and in amounts sufficiently large to fully substitute private credit. Section 5.2.shows that Bancóldex is not reducing interest rates importantly, and certainly not reducing rates beyond the opportunity costs of funds. Results in that section also suggest that Bancóldex credit does not fully substitute private credit.

22_{The effect on materials is insignificant (both statistically and economically) for small establishments in this}

(25)

to those that did not obtain Bancóldex loans. New participation models are estimated for the probability of receiving the respective type of loan. The restricted dataset used for columns marked FE+PSM in this case was constructed using a re-estimated matching exercise where inclusion in the restricted dataset is based on the probability of receiving a long-run Bancóldex loan in the upper panel, and the probability of receiving a short-run Bancóldex loan in the lower panel.

Our results show that long-run Bancóldex loans have a large positive impact on investment that is not observed when focusing on short-run loans. Not only are the coefficients for short-run Bancóldex loans on investment statistically insignificant, but they are also less than a third in magnitude of those for long-run loans. The opposite is true regarding the effect of Bancóldex on materials purchases; it is short run loans that most impact this outcome. Meanwhile, the effects on employment and output are positive and significant when considering both long-run and short-run loans, though the coefficients are slightly smaller for long-run loans.

Table 7 complements these findings by estimating the heterogeneous effects of long-

vs. short-run loans over time on materials, employment, and output.23_{The effects on}

materials purchases do not exhibit persistence for either type of loan—while in these results the effect on materials continues to be particularly large for short-run loans. For employment and output, the effect of long-run loans exhibits persistence over time, while the effect of short-run loans does not. These findings suggest that: 1) Bancóldex short-run loans play an important role in the provision of working capital but apparently not in the financing of investment projects, while its long-run loans are used to fund investments; and 2) at least part of the effect of long-run loans, initially reflected on investment, takes time to translate into complementary demand for labor and into greater output, but once it does that effect is more permanent in nature.

23_{Results on investment are also reported for consistency, but we abstain from emphasizing them on the}

(26)

Finally, Table 8 examines the robustness of our results to controlling for the lagged dependent variable rather than introducing establishment fixed effects as our approach to deal with potential selection concerns. Each vertical panel examines one of the four outcomes (employment, materials purchases, investment, output). Starting with the first column, and comparing it to results in Table 4, our results regarding the effect of Bancóldex loans on materials, employment and output are robust to this alternative estimation method. For employment and output this is true not only in terms of direction and statistical significance, but also in terms of the magnitude of the estimated coefficients. The estimated magnitude is sensible to controlling for the lagged dependent variable in the case of investment and materials purchases, both outcomes likely much more volatile than employment. For these outcome variables we continue to find a positive and significant effect, but now with a much larger magnitude, especially for investment. As previously pointed out, investment is known to exhibit a lumpy behavior. It is thus probably not surprising to find stark differences when approaching potential selection by including fixed plant effects or by including variable effects (in this particular case, the lagged dependent variable).

(27)

5.2. Bancóldex and credit conditions

We now move to the estimations of the effect of Bancóldex on the credit conditions faced by firms. That is, we estimate equation (1) but now setting the dependent variable to be one of several characteristics of a firm’s basket of loans from supervised financial intermediaries. These characteristics are: the average interest rate over the loans obtained in

the period; the average maturity specified for those loans;24_{the number of intermediaries}

from which the firm obtained loans over the period; and the amount of credit granted to the firm. In light of the conceptual framework in Section 3, we aim at understanding to

what extent Bancóldex credit is substituting private credit rather than complementing it.25

Because the set of data we use in this section is constructed from records on supervised credit operations, only firms receiving at least one loan from a supervised financial intermediaries over a given year are included in that particular year. This implies that in this section we only estimate the effect on credit conditions contingent on receiving at least one loan in the period. At the same time, recipient firms from all sectors are included. Even if we wanted to restrict the data to manufacturing firms for consistency with our results on performance, we would not be able to do so due to the lack of sector identifiers in the

credit data.26

Descriptive statistics for estimations relating the effect of Bancóldex on credit conditions are presented in Table 9. About 6% of all observations correspond to Bancóldex beneficiaries. The average interest rate is 21.7%; the average agreed maturity of a loan—expressed here in logs of days—is about 18 months; and while most firms have loans from a single financial intermediary, the range is wide, going up to 17 intermediaries for a single firm.

24_{That is, the maturity initially agreed upon. Firms in Colombia are able to repay a loan in advance of the}

pre-set maturity, though over the period of study advanced payment was sometimes penalized.

25_{As mentioned, the results we report in this section exclude year 2009, given the stark increase in loans}

classified as microcredit that occurred that year. Such increase implies a change in the composition of Bancóldex loans likely to have changed the type of beneficiaries; these compositional effects may end up captured by the results. Alternative results including year 2009 were also produced. They are not reported here in the interest of succinctness, but they lead to similar qualitative conclusions, in general with smaller magnitudes but similar levels of statistical significance.Results are available from authors upon request.

26_{The firm identifiers in this dataset are fictitious, so we are also unable to link firms to their sectors using}

(28)

As in the analysis of performance of the previous section, we conduct estimations using fixed effects—at the firm level in this case—and also estimations combining fixed effects with propensity score matching. Matching is done on the basis of year t-1 conditions of credit, as described in section 4. Figure 5 depicts the distribution of propensity scores in both the unrestricted and the restricted datasets. Restricting to beneficiaries and their nearest neighbors eliminates pre-existing differences in propensity scores. For individual outcomes of interest, pre-existing differences between treatment and control groups are also strongly reduced by the matching. Some remain statistically significant but their magnitudes is much smaller than that found below as a result of receiving a Bancóldex loan (Appendix A2).

Table 10 shows our results. Consistent with our findings of positive effects on firm performance, we find that Bancóldex beneficiaries see their total credit increased by almost 50%. They also see the average interest rates of their loans reduced, an indication that Bancóldex loans are making credit cheaper to the firm, and thus potentially displacing, at least partially, market credit. As clear from the conceptual framework presented in section 3, the degree to which Bancóldex is substituting market credit is important to understand the extent to which Bancóldex is easing credit constraints or credit rationing, rather than simply substituting credit that would be available though at a higher costs, and potentially expanding the demand for credit through this price effect.

There are several indications that Bancóldex is indeed not simply substituting market credit. On the one hand, the magnitude of the effect on the interest rate is not particularly

large: it represents a reduction of 2 pp, over an average level of over 21 pp.27_{Notice also}

that the use of Bancóldex is associated with an increase in the number of lenders from which the firm obtains loans of 0.4, which is more than half a standard deviation. Interestingly, the average maturity of loans also goes up by about 20%. Bancóldex seems to be actually widening the access of firms to private lenders, and at the same type providing access to a type of credit (long run) from which firms seem to have been most strongly rationed. This interpretation is also consistent with the fact that Bancóldex credit is not

27_{The average interest rate level of 21pp is likely more than 10 pp above the opportunity cost of funds to}

(29)

explicitly subsidized, and that the sources of funding that Bancóldex itself uses are not cheaper, probably even more expensive, than those of private banks, mainly for short-run funds. In the absence of explicit subsidies, the fall in average interest rates associated with Bancóldex credit probably reflects the expanded access to long run loans, for which Bancóldex may indeed be competitive in terms of the costs of funds.

Table 11 gets at the question of how Bancóldex affects market lending to firms from a different angle. It shows the results of estimating the effect of Bancóldex credit on the

amount of credit from sources other than Bancóldex. That is, we calculate total credit and the

number of lenders from which those loans come but excluding now all loans with Bancóldex funding. While a contraction in other sources of credit is indeed observed in the current year, its magnitude is far from implying a full substitution of market credit by Bancóldex. Moreover, such contraction lasts for a single period, and is then followed by an expansion in non-Bancóldex credit. The simple model presented in Section 3 cannot account for these dynamic effects. But, even in the absence of a tight interpretation in light of the model, these findings are at least suggestive that the persistent effects on performance found in the previous subsection are at least partly attributable to sustained credit that, despite being associated with having had a Bancóldex loan, is not coming solely from Bancóldex.

Table 12, presents robustness exercises including the lagged dependent variable as a regressor in lieu of fixed effects. Results for the effect of Bancóldex credit on overall credit conditions are remarkably similar to those shown in Table 10 (the baseline specification), in terms of sign, magnitude and significance. For credit from non-Bancóldex sources, however, controlling for lagged credit from non-Bancóldex sources changes the sign of the estimated contemporaneous effect of Bancóldex. The change of sign probably suggests that Bancóldex credit is going to firms previously rationed in their access to market credit, and that once this time-varying selection is accounted for, Bancóldex credit does not reduce the use of market credit, and may even increase it.

(30)

of lending to firms. As should be clear from the conceptual framework in section 3, Bancóldex could conceivably play the sole role of providing cheaper resources that firms would anyway find elsewhere and indeed use. If this were the case, credit would not be expanded: though firms would take-up Bancóldex-funded loans, they would use them to replace more costly loans from other sources. We do observe a reduction in the average interest rate faced by recipients of Bancóldex, but short lived and of a magnitude of less than a fifth of standard deviation. Meanwhile, firms do see their credit expanded after receiving Bancóldex, interact with new intermediaries, and do not fully substitute market credit for Bancóldex credit. Positive effects on credit are long-lived despite the interest rate effect disappearing.

6. Conclusions

Does public credit to firms alleviate credit constraints, or does it simply increase profits for firms that had full access to credit at higher costs? The answer depends on how constrained firms are in the absence of public credit, and on whether public credit is subsidized and efficiently allocated. We use information on a program of public lending in Colombia that, by design, is not subsidized and is allocated using the same criteria according to which private banks allocate funds to loan applicants. The impact of this program on firm performance and on the conditions under which firms receive credit, therefore, shed light on the degree to which firms are constrained in their ex-ante access to credit.

(31)

(32)

References

Armendáriz de Aghion, B. 1999. “Development Banking ” Journal of Development Economics.

58: 83-100.

Bancóldex. 2014. “Presentación institutional: banco de desarrollo empresarial ” March

2014. Downloaded from http://www.bancoldex.com/documentos/5081_Presentaci%C3%B3n_Institucion

al_Marzo_2014.pdf on August 22nd_2014.

Banerjee, A and E Duflo 2014 “Do firms want to borrow more? Testing Credit

Constraints using a Directed Lending Program” Forth Review of Economic Studies.

Barth, J , G Caprio Jr , and R Levine 1999 “Banking Systems Around the Globe: Do Regulation and Ownership Affect Performance and Stability?” In R E Litan and R Herring, eds., Brookings-Wharton Papers on Financial Services 2001. Washington, D.C.: Brookings Institution Press

Beck, T , and R Levine 2002 “Industry Growth and Capital Allocation: Does Having a

Market- or Bank-based System atter?” Journal of Financial Economics. 64: 147-180.

Carvalho, D. 2014 “The Real Effects of Government-owned Banks ” Journal of Finance

69(2).

Cole, Shawn. 2009. "Fixing Market Failures or Fixing Elections? Agricultural Credit in

India." American Economic Journal: Applied Economics: 1: 219–50.

Cooper, Russell and John Haltiwanger. 2006. "On the Nature of Capital Adjustment

Costs," Review of Economic Studies, 73(3): 611-633.

De Mel, S. McKenzie, D. and Woodruff, C. 2008. "Returns to Capital in Microenterprises:

Evidence from a Field Experiment," The Quarterly Journal of Economics, 123(4):

1329-1372.

Dewatripont, and E askin 1995 “Credit and efficiency in centralized and

decentralized economies,” Review of Economic Studies 62:541—55,

Dinç, S. 2005. “Politicians and Banks: Political Influences in Government-Owned Banks in

(33)

Doms, Mark and Tim Dunne. 1998. "Capital Adjustment Patterns in U.S. Manufacturing

Plants." Review of Economic Dynamics, 1: 409-429.

Galindo, A and A icco 2004 “Do State-owned banks Promote Growth? Cross-country

Evidence for anufacturing Industries ” Economics Letters: 84: 371-376.

La Porta, R , F Lopez de Silanes, and A Shleifer 2002 “Government Ownership of

Banks,” Journal of Finance 62: 265-302.

McKenzie, D. and C. Woodruff. 2008. "Experimental Evidence on Returns to Capital and

Access to Finance in Mexico," The World Bank Economic Review, 22(3), 457-482.

icco, A , U Panizza, and Yáñez 2007 “Bank ownership and performance. Does

politics matter?” Journal of Banking & Finance 31: 219–241

Sapienza, P 2004 “The Effects of Government Ownership on Bank Lending ” Journal of

Financial Economics 72: 357-384.

(34)

(35)

Figure 5: Distribution of Propensity Scores, Credit Analysis

0

5

10

15

20

P

rob

ab

ilit

y D

en

si

ty

F

un

ct

ion

0 .1 .2 .3 .4 Propensity Score

Control Treated

0

5

10

15

P

rob

ab

ilit

y D

en

si

ty

F

un

ct

ion

0 .1 .2 .3 .4 Propensity Score

Control Treated

All firms

(36)

Table 1. Bancóldex credit, total and intermediated by supervised financial system. 2004-2009 Total Micro firms Small firms Medium firms Large firms Total Micro firms Small firms Medium firms Large firms 2004 64.841 90,6% 6,6% 2,1% 0,7% 1.084.601 15% 20% 23% 42% 2005 69.824 90,3% 7,5% 1,7% 0,5% 1.007.385 19% 26% 21% 35% 2006 114.091 94,1% 4,8% 0,9% 0,2% 981.248 27% 32% 20% 21% 2007 133.187 94,6% 4,0% 1,2% 0,3% 1.693.248 16% 21% 22% 40% 2008 117.689 93,4% 4,6% 1,6% 0,4% 1.718.301 15% 22% 26% 37% 2009 149.710 94,6% 3,8% 1,3% 0,3% 1.539.698 17% 24% 27% 33% 2004 14.955 60,1% 28,0% 9,0% 3,0% 1.016.319 10% 21% 25% 45% 2005 17.323 61,6% 29,5% 6,9% 2,0% 920.869 12% 27% 23% 38% 2006 20.825 68,4% 25,6% 5,0% 0,9% 880.302 19% 35% 22% 24% 2007 19.746 64,0% 26,2% 7,9% 1,9% 1.571.607 10% 23% 24% 44% 2008 24.242 68,4% 21,7% 7,8% 2,2% 1.601.497 9% 23% 28% 39% 2009 87.377 90,8% 6,4% 2,2% 0,6% 1.470.883 13% 24% 28% 34%

Panel B. Through supervised financial intermediaries

Credit

Panel A. Total

(37)

Table 2: Bancóldex credit vs. all credit in supervised financial system

Year

Average

interest rate

(% )

Average term (in days) Average loan size (USD _thousand) Average interest rate (% ) Average term (in days) Average loan size (USD _thousand)

In credit value In number of relationships

2004 18,3 889 60,5 16,6 681 69,2 5,0% 4,4% 2005 17,0 933 50,2 15,5 802 51,2 3,7% 3,6% 2006 17,3 992 55,0 16,5 973 48,6 2,9% 3,3% 2007 22,6 1.039 51,3 19,7 1.002 117,4 6,9% 3,0% 2008 27,1 924 48,4 24,9 1.013 120,3 8,1% 3,2% 2009 25,2 892 41,0 30,3 652 21,4 5,9% 11,4% 2004 18,4 847 98,2 15,8 770 91,2 5,1% 5,5% 2005 17,9 889 98,4 14,9 882 64,1 3,6% 5,5% 2006 17,4 1.027 111,9 16,3 1.156 68,2 2,7% 4,4% 2007 21,3 1.157 97,6 16,6 1.233 173,8 6,7% 3,8% 2008 23,4 1.140 146,6 18,9 1.361 235,7 8,2% 5,1% 2009 17,9 1.127 163,3 13,9 1.195 143,9 5,5% 6,3% 2004 18,6 961 2,2 18,9 461 2,2 2,7% 2,8% 2005 16,5 988 2,1 17,7 611 2,2 1,9% 1,9% 2006 17,2 981 2,0 16,9 699 1,9 2,1% 2,2% 2007 23,7 948 1,9 24,6 660 1,6 1,8% 2,2% 2008 28,7 841 2,0 30,3 704 2,5 3,0% 2,4% 2009 27,5 825 1,9 32,9 573 1,1 7,7% 13,0%

Panel C: Microcredit Source: Superfinanciera and authors’ calculations. Monetary values converted to 2009 pesos using the CPI and then to dollars at the December 2009 peso/dollar exchange rate. Numbers refer to loan operations through financial intermediaries.

Bancóldex credit relationships

All credit relationships

Panel A: All

Bancóldex participation

(38)

N Mean Std. Dev.

Dummy Bancoldex=1 22.386 0,11 0,31

Log employment 22.382 3,37 1,21

Log input consumption 22.386 6,76 5,47

Log investment 19.519 7,11 5,39

Log output 22.386 14,59 1,72

Dummy multiestablishment firm =1 22.386 0,08 0,27

Age 22.386 25,79 14,25

Dummy for positive interest payments

on financial obligations 22.386 0,68 0,47

Dummy Bancóldex long run loan =1 21.522 0,07 0,26 Dumm Bancóldex short run loan =1 20.597 0,03 0,17

Bancóldex loan value 22.386 1,30 3,76